Do you think this should be backported to 2.2.2 as well?
I already backported the part that's arguably "a bugfix" (the part that would have solved Kevin's severe speed problem, had 2.2.1 not had a different bug that prevented the dramatic slowdown he saw in CVS Python).
All that remains is new optimizations, and those can't be sold as a bugfix. Large (in the sense of lines of code) changes to gc really need to go thru lots of testing too. The optimizations involve some new algorithms, not just tweaking the former ones.
So, no. If it were-- or evolves into --a dramatic speedup, maybe.